Research On Approaches To Predicate Annotation Based On Semantic

How to transfer natural language texts into structured data is one of the research focuses in the field of Artificial Intelligence. Through making semantic annotation on natural language texts, gaps between data in natural language and structured data could be bridged, so that machines could mine knowledge and reason on this natural language data. Semantic annotation consists of entity annotation which deal with the subject and object and predicate annotation which deal with the verb in the subject-verb-object structure. In contrast to the massive research on the entity annotation, predicate annotation is a much less popular area in the common domain.Base on those facts, this thesis is focused on the research of how to make semantic extension on verb phrase, how to construct verb phrase library from the extension result, and make predicate annotation based on the library. More specifically, the three aspects are involved:1) Research on verb phrases semantic extensions. With the common domain knowledge base regarded as the knowledge background, this thesis extracts verb phrases from natural language texts, and extends the semantic of these verb phrases using the paths of binary relations in the knowledge base.2) Research on the construction of the verb phrases library. Through making an assessment of the semantic extension in the first step, this thesis ranks these extensions, and constructs a phrases library based on that.3) Research on the semantic annotation based on the phrases library. Both standard method and supplement methods are tried and assessed.As a conclusion, inspired by the former research, this thesis focuses on approaches to making verb phrase semantic extension, and building verb phrases library based on the extension results. Furthermore, an algorithm of predicate annotation is proposed and proved to be effective.